Expanding the usefulness of the serializers is something that we would like to address. However, it's not a trivial problem, and it will take some serious design work.

— Russell Keith-Magee, Django users group

Serializers allow complex data such as querysets and model instances to be converted to native Python datatypes that can then be easily rendered into JSON, XML or other content types. Serializers also provide deserialization, allowing parsed data to be converted back into complex types, after first validating the incoming data.

REST framework's serializers work very similarly to Django's Form and ModelForm classes. It provides a Serializer class which gives you a powerful, generic way to control the output of your responses, as well as a ModelSerializer class which provides a useful shortcut for creating serializers that deal with model instances and querysets.

Let's start by creating a simple object we can use for example purposes:

class Comment(object):
    def __init__(self, email, content, created=None):
        self.email = email
        self.content = content
        self.created = created or datetime.datetime.now()

comment = Comment(email='[email protected]', content='foo bar')

We'll declare a serializer that we can use to serialize and deserialize Comment objects.

Declaring a serializer looks very similar to declaring a form:

from rest_framework import serializers

class CommentSerializer(serializers.Serializer):
    email = serializers.EmailField()
    content = serializers.CharField(max_length=200)
    created = serializers.DateTimeField()

    def restore_object(self, attrs, instance=None):
        """
        Given a dictionary of deserialized field values, either update
        an existing model instance, or create a new model instance.
        """
        if instance is not None:
            instance.email = attrs.get('email', instance.email)
            instance.content = attrs.get('content', instance.content)
            instance.created = attrs.get('created', instance.created)
            return instance
        return Comment(**attrs) 

The first part of serializer class defines the fields that get serialized/deserialized. The restore_object method defines how fully fledged instances get created when deserializing data.

The restore_object method is optional, and is only required if we want our serializer to support deserialization into fully fledged object instances. If we don't define this method, then deserializing data will simply return a dictionary of items.

We can now use CommentSerializer to serialize a comment, or list of comments. Again, using the Serializer class looks a lot like using a Form class.

serializer = CommentSerializer(comment)
serializer.data
# {'email': u'[email protected]', 'content': u'foo bar', 'created': datetime.datetime(2012, 8, 22, 16, 20, 9, 822774)}

At this point we've translated the model instance into Python native datatypes. To finalise the serialization process we render the data into json.

from rest_framework.renderers import JSONRenderer

json = JSONRenderer().render(serializer.data)
json
# '{"email": "[email protected]", "content": "foo bar", "created": "2012-08-22T16:20:09.822"}'

Sometimes when serializing objects, you may not want to represent everything exactly the way it is in your model.

If you need to customize the serialized value of a particular field, you can do this by creating a transform_<fieldname> method. For example if you needed to render some markdown from a text field:

description = serializers.CharField()
description_html = serializers.CharField(source='description', read_only=True)

def transform_description_html(self, obj, value):
    from django.contrib.markup.templatetags.markup import markdown
    return markdown(value)

These methods are essentially the reverse of validate_<fieldname> (see Validation below.)

Deserialization is similar. First we parse a stream into Python native datatypes...

from StringIO import StringIO
from rest_framework.parsers import JSONParser

stream = StringIO(json)
data = JSONParser().parse(stream)

...then we restore those native datatypes into a fully populated object instance.

serializer = CommentSerializer(data=data)
serializer.is_valid()
# True
serializer.object
# <Comment object at 0x10633b2d0>

When deserializing data, we can either create a new instance, or update an existing instance.

serializer = CommentSerializer(data=data)           # Create new instance
serializer = CommentSerializer(comment, data=data)  # Update `comment`

By default, serializers must be passed values for all required fields or they will throw validation errors. You can use the partial argument in order to allow partial updates.

serializer = CommentSerializer(comment, data={'content': u'foo bar'}, partial=True)  # Update `comment` with partial data

When deserializing data, you always need to call is_valid() before attempting to access the deserialized object. If any validation errors occur, the .errors property will contain a dictionary representing the resulting error messages. For example:

serializer = CommentSerializer(data={'email': 'foobar', 'content': 'baz'})
serializer.is_valid()
# False
serializer.errors
# {'email': [u'Enter a valid e-mail address.'], 'created': [u'This field is required.']}

Each key in the dictionary will be the field name, and the values will be lists of strings of any error messages corresponding to that field. The non_field_errors key may also be present, and will list any general validation errors.

When deserializing a list of items, errors will be returned as a list of dictionaries representing each of the deserialized items.

You can specify custom field-level validation by adding .validate_<fieldname> methods to your Serializer subclass. These are analogous to .clean_<fieldname> methods on Django forms, but accept slightly different arguments.

They take a dictionary of deserialized attributes as a first argument, and the field name in that dictionary as a second argument (which will be either the name of the field or the value of the source argument to the field, if one was provided).

Your validate_<fieldname> methods should either just return the attrs dictionary or raise a ValidationError. For example:

from rest_framework import serializers

class BlogPostSerializer(serializers.Serializer):
    title = serializers.CharField(max_length=100)
    content = serializers.CharField()

    def validate_title(self, attrs, source):
        """
        Check that the blog post is about Django.
        """
        value = attrs[source]
        if "django" not in value.lower():
            raise serializers.ValidationError("Blog post is not about Django")
        return attrs

To do any other validation that requires access to multiple fields, add a method called .validate() to your Serializer subclass. This method takes a single argument, which is the attrs dictionary. It should raise a ValidationError if necessary, or just return attrs. For example:

from rest_framework import serializers

class EventSerializer(serializers.Serializer):
    description = serializers.CharField(max_length=100)
    start = serializers.DateTimeField()
    finish = serializers.DateTimeField()

    def validate(self, attrs):
        """
        Check that the start is before the stop.
        """
        if attrs['start'] > attrs['finish']:
            raise serializers.ValidationError("finish must occur after start")
        return attrs

To save the deserialized objects created by a serializer, call the .save() method:

if serializer.is_valid():
    serializer.save()

The default behavior of the method is to simply call .save() on the deserialized object instance. You can override the default save behaviour by overriding the .save_object(obj) method on the serializer class.

The generic views provided by REST framework call the .save() method when updating or creating entities.

The previous examples are fine for dealing with objects that only have simple datatypes, but sometimes we also need to be able to represent more complex objects, where some of the attributes of an object might not be simple datatypes such as strings, dates or integers.

The Serializer class is itself a type of Field, and can be used to represent relationships where one object type is nested inside another.

class UserSerializer(serializers.Serializer):
    email = serializers.EmailField()
    username = serializers.CharField(max_length=100)

class CommentSerializer(serializers.Serializer):
    user = UserSerializer()
    content = serializers.CharField(max_length=200)
    created = serializers.DateTimeField()

If a nested representation may optionally accept the None value you should pass the required=False flag to the nested serializer.

class CommentSerializer(serializers.Serializer):
    user = UserSerializer(required=False)  # May be an anonymous user.
    content = serializers.CharField(max_length=200)
    created = serializers.DateTimeField()

Similarly if a nested representation should be a list of items, you should pass the many=True flag to the nested serialized.

class CommentSerializer(serializers.Serializer):
    user = UserSerializer(required=False)
    edits = EditItemSerializer(many=True)  # A nested list of 'edit' items.
    content = serializers.CharField(max_length=200)
    created = serializers.DateTimeField()

Validation of nested objects will work the same as before. Errors with nested objects will be nested under the field name of the nested object.

serializer = CommentSerializer(data={'user': {'email': 'foobar', 'username': 'doe'}, 'content': 'baz'})
serializer.is_valid()
# False
serializer.errors
# {'user': {'email': [u'Enter a valid e-mail address.']}, 'created': [u'This field is required.']}

The Serializer class can also handle serializing or deserializing lists of objects.

To serialize a queryset or list of objects instead of a single object instance, you should pass the many=True flag when instantiating the serializer. You can then pass a queryset or list of objects to be serialized.

queryset = Book.objects.all()
serializer = BookSerializer(queryset, many=True)
serializer.data
# [
#     {'id': 0, 'title': 'The electric kool-aid acid test', 'author': 'Tom Wolfe'},
#     {'id': 1, 'title': 'If this is a man', 'author': 'Primo Levi'},
#     {'id': 2, 'title': 'The wind-up bird chronicle', 'author': 'Haruki Murakami'}
# ]

To deserialize a list of object data, and create multiple object instances in a single pass, you should also set the many=True flag, and pass a list of data to be deserialized.

This allows you to write views that create multiple items when a POST request is made.

For example:

data = [
    {'title': 'The bell jar', 'author': 'Sylvia Plath'},
    {'title': 'For whom the bell tolls', 'author': 'Ernest Hemingway'}
]
serializer = BookSerializer(data=data, many=True)
serializer.is_valid()
# True
serializer.save()  # `.save()` will be called on each deserialized instance

You can also deserialize a list of objects as part of a bulk update of multiple existing items. In this case you need to supply both an existing list or queryset of items, as well as a list of data to update those items with.

This allows you to write views that update or create multiple items when a PUT request is made.

# Capitalizing the titles of the books
queryset = Book.objects.all()
data = [
    {'id': 3, 'title': 'The Bell Jar', 'author': 'Sylvia Plath'},
    {'id': 4, 'title': 'For Whom the Bell Tolls', 'author': 'Ernest Hemingway'}
]
serializer = BookSerializer(queryset, data=data, many=True)
serializer.is_valid()
# True
serializer.save()  # `.save()` will be called on each updated or newly created instance.

By default bulk updates will be limited to updating instances that already exist in the provided queryset.

When performing a bulk update you may want to allow new items to be created, and missing items to be deleted. To do so, pass allow_add_remove=True to the serializer.

serializer = BookSerializer(queryset, data=data, many=True, allow_add_remove=True)
serializer.is_valid()
# True
serializer.save()  # `.save()` will be called on updated or newly created instances.
                   # `.delete()` will be called on any other items in the `queryset`.

Passing allow_add_remove=True ensures that any update operations will completely overwrite the existing queryset, rather than simply updating existing objects.

Performing a bulk update is slightly more complicated than performing a bulk creation, because the serializer needs a way to determine how the items in the incoming data should be matched against the existing object instances.

By default the serializer class will use the id key on the incoming data to determine the canonical identity of an object. If you need to change this behavior you should override the get_identity method on the Serializer class. For example:

class AccountSerializer(serializers.Serializer):
    slug = serializers.CharField(max_length=100)
    created = serializers.DateTimeField()
    ...  # Various other fields
    
    def get_identity(self, data):
        """
        This hook is required for bulk update.
        We need to override the default, to use the slug as the identity.
        
        Note that the data has not yet been validated at this point,
        so we need to deal gracefully with incorrect datatypes.
        """
        try:
            return data.get('slug', None)
        except AttributeError:
            return None

To map the incoming data items to their corresponding object instances, the .get_identity() method will be called both against the incoming data, and against the serialized representation of the existing objects.

There are some cases where you need to provide extra context to the serializer in addition to the object being serialized. One common case is if you're using a serializer that includes hyperlinked relations, which requires the serializer to have access to the current request so that it can properly generate fully qualified URLs.

You can provide arbitrary additional context by passing a context argument when instantiating the serializer. For example:

serializer = AccountSerializer(account, context={'request': request})
serializer.data
# {'id': 6, 'owner': u'denvercoder9', 'created': datetime.datetime(2013, 2, 12, 09, 44, 56, 678870), 'details': 'http://example.com/accounts/6/details'}

The context dictionary can be used within any serializer field logic, such as a custom .to_native() method, by accessing the self.context attribute.

Often you'll want serializer classes that map closely to model definitions. The ModelSerializer class lets you automatically create a Serializer class with fields that correspond to the Model fields.

class AccountSerializer(serializers.ModelSerializer):
    class Meta:
        model = Account

By default, all the model fields on the class will be mapped to corresponding serializer fields.

Any relationships such as foreign keys on the model will be mapped to PrimaryKeyRelatedField. Other models fields will be mapped to a corresponding serializer field.

Note: When validation is applied to a ModelSerializer, both the serializer fields, and their corresponding model fields must correctly validate. If you have optional fields on your model, make sure to correctly set blank=True on the model field, as well as setting required=False on the serializer field.

If you only want a subset of the default fields to be used in a model serializer, you can do so using fields or exclude options, just as you would with a ModelForm.

For example:

class AccountSerializer(serializers.ModelSerializer):
    class Meta:
        model = Account
        fields = ('id', 'account_name', 'users', 'created')

The default ModelSerializer uses primary keys for relationships, but you can also easily generate nested representations using the depth option:

class AccountSerializer(serializers.ModelSerializer):
    class Meta:
        model = Account
        fields = ('id', 'account_name', 'users', 'created')
        depth = 1

The depth option should be set to an integer value that indicates the depth of relationships that should be traversed before reverting to a flat representation.

If you want to customize the way the serialization is done (e.g. using allow_add_remove) you'll need to define the field yourself.

You may wish to specify multiple fields as read-only. Instead of adding each field explicitly with the read_only=True attribute, you may use the read_only_fields Meta option, like so:

class AccountSerializer(serializers.ModelSerializer):
    class Meta:
        model = Account
        fields = ('id', 'account_name', 'users', 'created')
        read_only_fields = ('account_name',)

Model fields which have editable=False set, and AutoField fields will be set to read-only by default, and do not need to be added to the read_only_fields option.

You may wish to specify multiple fields as write-only. Instead of adding each field explicitly with the write_only=True attribute, you may use the write_only_fields Meta option, like so:

class CreateUserSerializer(serializers.ModelSerializer):
    class Meta:
        model = User
        fields = ('email', 'username', 'password')
        write_only_fields = ('password',)  # Note: Password field is write-only

    def restore_object(self, attrs, instance=None):
        """
        Instantiate a new User instance.
        """
        assert instance is None, 'Cannot update users with CreateUserSerializer'                                
        user = User(email=attrs['email'], username=attrs['username'])
        user.set_password(attrs['password'])
        return user

You can add extra fields to a ModelSerializer or override the default fields by declaring fields on the class, just as you would for a Serializer class.

class AccountSerializer(serializers.ModelSerializer):
    url = serializers.CharField(source='get_absolute_url', read_only=True)
    groups = serializers.PrimaryKeyRelatedField(many=True)

    class Meta:
        model = Account

Extra fields can correspond to any property or callable on the model.

When serializing model instances, there are a number of different ways you might choose to represent relationships. The default representation for ModelSerializer is to use the primary keys of the related instances.

Alternative representations include serializing using hyperlinks, serializing complete nested representations, or serializing with a custom representation.

For full details see the serializer relations documentation.

## Inheritance of the 'Meta' class

The inner Meta class on serializers is not inherited from parent classes by default. This is the same behaviour as with Django's Model and ModelForm classes. If you want the Meta class to inherit from a parent class you must do so explicitly. For example:

class AccountSerializer(MyBaseSerializer):
    class Meta(MyBaseSerializer.Meta):
        model = Account

Typically we would recommend not using inheritance on inner Meta classes, but instead declaring all options explicitly.

The HyperlinkedModelSerializer class is similar to the ModelSerializer class except that it uses hyperlinks to represent relationships, rather than primary keys.

By default the serializer will include a url field instead of a primary key field.

The url field will be represented using a HyperlinkedIdentityField serializer field, and any relationships on the model will be represented using a HyperlinkedRelatedField serializer field.

You can explicitly include the primary key by adding it to the fields option, for example:

class AccountSerializer(serializers.HyperlinkedModelSerializer):
    class Meta:
        model = Account
        fields = ('url', 'id', 'account_name', 'users', 'created')

There needs to be a way of determining which views should be used for hyperlinking to model instances.

By default hyperlinks are expected to correspond to a view name that matches the style '{model_name}-detail', and looks up the instance by a pk keyword argument.

You can change the field that is used for object lookups by setting the lookup_field option. The value of this option should correspond both with a kwarg in the URL conf, and with a field on the model. For example:

class AccountSerializer(serializers.HyperlinkedModelSerializer):
    class Meta:
        model = Account
        fields = ('url', 'account_name', 'users', 'created')
        lookup_field = 'slug'

Note that the lookup_field will be used as the default on all hyperlinked fields, including both the URL identity, and any hyperlinked relationships.

For more specific requirements such as specifying a different lookup for each field, you'll want to set the fields on the serializer explicitly. For example:

class AccountSerializer(serializers.HyperlinkedModelSerializer):
    url = serializers.HyperlinkedIdentityField(
        view_name='account_detail',
        lookup_field='account_name'
    )
    users = serializers.HyperlinkedRelatedField(
        view_name='user-detail',
        lookup_field='username',
        many=True,
        read_only=True
    )

    class Meta:
        model = Account
        fields = ('url', 'account_name', 'users', 'created')

The name of the URL field defaults to 'url'. You can override this globally, by using the URL_FIELD_NAME setting.

You can also override this on a per-serializer basis by using the url_field_name option on the serializer, like so:

class AccountSerializer(serializers.HyperlinkedModelSerializer):
    class Meta:
        model = Account
        fields = ('account_url', 'account_name', 'users', 'created')
        url_field_name = 'account_url'

Note: The generic view implementations normally generate a Location header in response to successful POST requests. Serializers using url_field_name option will not have this header automatically included by the view. If you need to do so you will ned to also override the view's get_success_headers() method.

You can also override the URL field's view name and lookup field without overriding the field explicitly, by using the view_name and lookup_field options, like so:

class AccountSerializer(serializers.HyperlinkedModelSerializer):
    class Meta:
        model = Account
        fields = ('account_url', 'account_name', 'users', 'created')
        view_name = 'account_detail'
        lookup_field='account_name'

You can create customized subclasses of ModelSerializer or HyperlinkedModelSerializer that use a different set of default fields.

Doing so should be considered advanced usage, and will only be needed if you have some particular serializer requirements that you often need to repeat.

Once a serializer has been initialized, the dictionary of fields that are set on the serializer may be accessed using the .fields attribute. Accessing and modifying this attribute allows you to dynamically modify the serializer.

Modifying the fields argument directly allows you to do interesting things such as changing the arguments on serializer fields at runtime, rather than at the point of declaring the serializer.

For example, if you wanted to be able to set which fields should be used by a serializer at the point of initializing it, you could create a serializer class like so:

class DynamicFieldsModelSerializer(serializers.ModelSerializer):
    """
    A ModelSerializer that takes an additional `fields` argument that
    controls which fields should be displayed.
    """

    def __init__(self, *args, **kwargs):
        # Don't pass the 'fields' arg up to the superclass
        fields = kwargs.pop('fields', None)
        
        # Instantiate the superclass normally
        super(DynamicFieldsModelSerializer, self).__init__(*args, **kwargs)

        if fields:
            # Drop any fields that are not specified in the `fields` argument.
            allowed = set(fields)
            existing = set(self.fields.keys())
            for field_name in existing - allowed:
                self.fields.pop(field_name)

This would then allow you to do the following:

>>> class UserSerializer(DynamicFieldsModelSerializer):
>>>     class Meta:
>>>         model = User
>>>         fields = ('id', 'username', 'email')
>>>
>>> print UserSerializer(user)
{'id': 2, 'username': 'jonwatts', 'email': '[email protected]'}
>>>
>>> print UserSerializer(user, fields=('id', 'email'))
{'id': 2, 'email': '[email protected]'}

The field_mapping attribute is a dictionary that maps model classes to serializer classes. Overriding the attribute will let you set a different set of default serializer classes.

For more advanced customization than simply changing the default serializer class you can override various get_<field_type>_field methods. Doing so will allow you to customize the arguments that each serializer field is initialized with. Each of these methods may either return a field or serializer instance, or None.

Signature: .get_pk_field(self, model_field)

Returns the field instance that should be used to represent the pk field.

Signature: .get_nested_field(self, model_field, related_model, to_many)

Returns the field instance that should be used to represent a related field when depth is specified as being non-zero.

Note that the model_field argument will be None for reverse relationships. The related_model argument will be the model class for the target of the field. The to_many argument will be a boolean indicating if this is a to-one or to-many relationship.

Signature: .get_related_field(self, model_field, related_model, to_many)

Returns the field instance that should be used to represent a related field when depth is not specified, or when nested representations are being used and the depth reaches zero.

Note that the model_field argument will be None for reverse relationships. The related_model argument will be the model class for the target of the field. The to_many argument will be a boolean indicating if this is a to-one or to-many relationship.

Signature: .get_field(self, model_field)

Returns the field instance that should be used for non-relational, non-pk fields.

The following custom model serializer could be used as a base class for model serializers that should always exclude the pk by default.

class NoPKModelSerializer(serializers.ModelSerializer):
    def get_pk_field(self, model_field):
        return None

The following third party packages are also available.

The django-rest-framework-mongoengine package provides a MongoEngineModelSerializer serializer class that supports using MongoDB as the storage layer for Django REST framework.

The django-rest-framework-gis package provides a GeoFeatureModelSerializer serializer class that supports GeoJSON both for read and write operations.

The django-rest-framework-hstore package provides an HStoreSerializer to support django-hstore DictionaryField model field and its schema-mode feature.